On 08/15/2012 05:09 PM, Michal Hocko wrote:
> On Wed 15-08-12 13:42:24, Glauber Costa wrote:
> [...]
>>>> +
>>>> + ret = 0;
>>>> +
>>>> + if (!memcg)
>>>> + return ret;
>>>> +
>>>> + _memcg = memcg;
>>>> + ret = __mem_cgroup_try_charge(NULL, gfp, delta / PAGE_SIZE,
>>>> + &_memcg, may_oom);
>>>
>>> This is really dangerous because atomic allocation which seem to be
>>> possible could result in deadlocks because of the reclaim.
>>
>> Can you elaborate on how this would happen?
>
> Say you have an atomic allocation and we hit the limit so we get either
> to reclaim which can sleep or to oom which can sleep as well (depending
> on the oom_control).
>
I see now, you seem to be right.
How about we change the following code in mem_cgroup_do_charge:
if (gfp_mask & __GFP_NORETRY)
return CHARGE_NOMEM;
to:
if ((gfp_mask & __GFP_NORETRY) || (gfp_mask & __GFP_ATOMIC))
return CHARGE_NOMEM;
?
Would this take care of the issue ?
--
To unsubscribe from this list: send the line "unsubscribe cgroups" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
